Environmental influences on the pace of brain development
Stress and resilience
Childhood socio-economic status (SES), a measure of the availability of material and social resources, is one of the strongest predictors of lifelong well-being. Here we review evidence that experiences associated with childhood SES affect not only the outcome but also the pace of brain development. We argue that higher childhood SES is associated with protracted structural brain development and a prolonged trajectory of functional network segregation, ultimately leading to more efficient cortical networks in adulthood. We hypothesize that greater exposure to chronic stress accelerates brain maturation, whereas greater access to novel positive experiences decelerates maturation. We discuss the impact of variation in the pace of brain development on plasticity and learning. We provide a generative theoretical framework to catalyse future basic science and translational research on environmental influences on brain development.
Children’s early experiences are associated with important later-life outcomes, including their earnings 1 , educational attainment 2 , physical well-being 3 and mental health 4 . How are children’s experiences embedded in their developing brains to broaden, or constrain, their opportunities to live happy and healthy lives? Much of what we know about links between early experiences and adult outcomes has come from research on socio-economic status (SES). A multidimensional construct, SES is typically measured at the household level (for example, parental income, education or occupation) or the neighbourhood level (for instance, neighbourhood crime rate, poverty levels or median income). Higher SES is associated with lower exposure to stress, and with greater access to cognitive enrichment, such as high-quality education, child-directed language, books and toys. Variation in childhood SES has been associated with variation in measures of brain structure and function 5 , 6 , 7 , 8 . However, surprisingly little is known about whether and how experiences associated with childhood SES affect the trajectory of brain maturation.
Here, we synthesize evidence that experiences associated with childhood SES affect not only the outcome, but also the pace of brain development, and consider the implications of early brain development for plasticity in childhood. We focus on whole-brain cortical measures of structure and function because, as a broad and multidimensional construct, SES probably exerts effects on a complex constellation of brain regions and their connections. We highlight the few longitudinal studies on SES and brain development but, because these studies are rare, we also draw on cross-sectional studies of relationships between SES and brain structure and function across development 9 . We consider how experiences, including stress, cognitive enrichment and environmental variability, influence brain maturation and plasticity. We close by outlining promising future directions for research on how children’s early experiences lead to disparities in later-life outcomes.
Structural brain development
Cortical thickness increases in the prenatal and immediate postnatal period, driven by dendritic and axonal growth as well as synaptogenesis 10 . Peak synaptic density and peak cortical thickness are reached at different times across the brain, with sensory regions showing faster development and earlier peaks, and association regions showing slower developmental trajectories 11 , 12 (Fig. 1 ). The cortex thickens before 2 years of age, before undergoing widespread thinning across a protracted period starting between 2 and 5 years of age, and continuing through adolescence and early adulthood. Thinning is attributed to both regressive (synaptic pruning) and progressive (myelination) processes 13 , 14 . In adulthood, a thicker cortex is associated with larger, more complex pyramidal neurons 15 . Cortical surface area increases during childhood and into early adolescence, with the greatest increases occurring first in sensory areas, and latest in association areas 16 , 17 .
Fig. 1: Associations between socio-economic status and cortical thickness.
Trajectories shown in light and dark blue are conceptual, based on findings interpolated across multiple studies. Horizontal grey lines represent the age ranges of individual studies, as shown on the horizontal axis. Brain regions shown in blue indicate negative relationships between socio-economic status (SES) and cortical thickness (ref. 21 corresponds to grey line 1). Brain regions shown in red indicate positive relationships between SES and cortical thickness (grey line 2, ref. 19 ; grey line 3, ref. 8 ; grey line 4, ref. 22 ; grey line 5, ref. 18 ; grey line 6, ref. 25 ; grey line 7, ref. 107 ; grey line 8, ref. 36 ; grey line 9, ref. 20 ; grey line 10, ref. 24 ; grey line 11, ref. 26 ). These curves are consistent with more modest main effects of SES on cortical thickness when averaging is done across large age ranges than when small age ranges are focused upon. The inset shows a schematic of potential cellular underpinnings of cortical thickness as measured by MRI: glial number and size, neuron number and size, synaptic complexity and myelination 14 , 15 , 16 . Cells are enlarged relative to cortical thickness to show detail. Brain image corresponding to grey line 1 adapted with permission from ref. 21 , OUP. Brain image corresponding to grey line 2 adapted with permission from ref. 19 , CC BY 4.0 ( https://creativecommons.org/licenses/by/4.0/ ). Brain image corresponding to grey line 3 adapted with permission from ref. 8 , Sage Publishing. Brain image corresponding to grey line 4 adapted with permission from ref. 22 , CC BY 4.0 ( https://creativecommons.org/licenses/by/4.0/ ).
Full size image
Children and adolescents from higher-SES environments generally have thicker cortex than those from lower-SES environments 8 , 18 , 19 , 20 , but there is evidence that relationships between SES and cortical thickness vary with age (Fig. 1 ). In the first postnatal year, when the cortex rapidly thickens, higher paternal education is associated with thinner cortex, particularly in the frontal lobes 21 . This pattern is suggestive of more prolonged maturational processes in infants from higher-SES backgrounds. Later in development, in youth aged 3–20 years, SES moderates the negative relationship between age and cortical thickness such that youth from lower-SES backgrounds show a steeper curvilinear decrease in cortical thickness at a younger age than do youth from higher-SES backgrounds 22 , 23 . Adolescents aged 12–18 years in low-income households show a steeper curvilinear relationship between age and cortical thickness than do adolescents in high-income households 24 . For females, but not males, in low-income households, living in high-inequality neighbourhoods is again associated with a steeper negative relationship between age and cortical thickness 24 . This evidence is consistent with the hypothesis that lower SES is associated with accelerated cortical thinning throughout childhood and adolescence. However, not all findings align with this hypothesis. Two recent studies examined youth aged 5–25 years 25 and 14–19 years 26 and did not find that SES moderated relationships between age and cortical thickness, although the former study reported positive correlations between SES and cortical thickness. However, examining a large age range such as 5–25 years might obscure interaction effects that vary over the course of development, and SES-related variability in the rate of cortical thinning during late adolescence when thinning has slowed may be minimal (Fig. 1 ). In addition, neither study examined non-linear relationships between age and cortical thickness moderated by SES.
Fewer studies have examined associations between SES and cortical surface area development. In infancy, surface area is not related to parental education or income 21 . In late childhood and adolescence, however, higher SES is associated with greater surface area 25 , 26 , 27 . In an analysis of the Pediatric Imaging, Neurocognition, and Genetics (PING) dataset, researchers applied sample weights to structural brain imaging data collected from children aged 3–18 years to create a ‘weighted sample’ approximating the distribution of SES, race/ethnicity and sex in the US population. When the researchers used the weighted sample to examine associations between surface area and age, the surface area peak shifted earlier as compared with the unweighted sample, consistent with an interpretation of earlier or faster brain maturation in children from lower-SES backgrounds, who were under-represented in the original sample 28 . In a recent longitudinal study of adolescents, higher SES was associated with a smaller decline in total surface area between 14 and 19 years of age 26 .
The cellular processes that underlie cortical thickness and surface area measures obtained with MRI are still under active investigation. As noted already, cortical thickness is positively associated with synaptic density, and is negatively associated with myelination 14 , 15 . One possibility is that experiences associated with low SES drive earlier curtailment of synaptic proliferation and a subsequently decreased range for optimal synaptic pruning and wiring of functional networks. Computational models of synaptic proliferation suggest that synaptic overgrowth and then pruning of weak synapses maximizes network performance, given the metabolic constraints of the brain 29 . In biologically motivated models of network development, delaying synaptogenesis in higher-order layers of a network leads to greater energy efficiency and faster learning after development 30 . Moreover, networks with more initial connections are better able to learn than networks with fewer initial connections 31 . Computational models of synaptic proliferation and subsequent pruning early in development have identified a trade-off between rapid development, which enables earlier independence and less parental input, and optimal adult neural performance 32 . SES-associated differences in early synaptic proliferation would affect the development of functional connectivity, which we examine in the following section.
Functional network development
A key goal of brain development is to establish efficient, specialized cortical systems. Functional activation of specific systems can be studied by imaging individuals performing well-designed tasks, but SES-associated differences in task accuracy and the interpretation of stimuli can affect conclusions about the underlying anatomy 33 . By contrast, data collected when participants relax inside the scanner — that is, resting-state functional MRI (rs-fMRI) data — can be used to study all systems simultaneously without task confounds 34 . Components of a functional system show statistically similar patterns of fluctuations in blood oxygenation, commonly referred to as functional connectivity 35 .
Resting-state analyses have generated conflicting answers to the question of whether higher SES is associated with faster functional maturation. One compelling study integrated grey and white matter structure with regional rs-fMRI measures to develop a model to classify individuals’ ages. It was found that individuals aged 8–22 years from lower-SES backgrounds were more likely to be classified as adults than their higher-SES counterparts 36 . Other rs-fMRI studies also suggest that lower SES is associated with faster functional development: in youth aged 6–17 years, lower SES was associated with weaker connectivity in corticostriatal connections that typically showed decreases in strength with age over development 37 , 38 . However, some studies have found the opposite pattern: higher SES has been associated with greater functional connectivity between limbic regions that typically show age-related increases in functional connectivity over development 39 , 40 , 41 . These studies largely examined patterns of regional metrics or connectivity between specific sets of regions rather than testing for broad effects of SES on the pace of network development throughout the brain. However, region-to-region connectivity can be strengthened by repeated co-activation, just as cells that fire together will wire together. Therefore, it is difficult to infer broad developmental processes from examining links between specific regions 42 .
Newer approaches to analysing rs-fMRI data are computationally better suited to test the hypothesis that higher childhood SES is associated with protracted development of functional networks across the entire cortex. A network science approach, in particular, represents the brain as a collection of nodes (regions) and edges (connections), enabling us to address the whole-brain pattern of connectivity 43 , 44 . The resulting network architecture can then be quantitatively characterized with use of tools from graph theory to identify key properties relevant to maturation 45 . Two such properties are segregation and integration, both of which change during development 46 . Segregation quantifies the presence of groups or subnetworks of densely interconnected nodes in a network, whereas integration assesses the extent to which information can be rapidly combined from distributed regions 43 . Integration has a distinct meaning when one is interpreting diffusion data compared with when one is interpreting functional data 47 (Box 1 ). Together, integration and segregation constitute the unique property of small-worldness found in adult brain networks: the perhaps counterintuitive presence of high levels of both segregation and integration at many different scales (see ref. 48 for a recent review). Given the associations between functional network segregation at rest and cognitive abilities 35 , 49 , and that most research on SES and functional network development has examined segregation rather than integration, we focus specifically here on measures of functional network segregation.
Segregation in brain networks changes markedly over development, and can be measured at several scales. One measure of segregation at the nodal level is the clustering coefficient, which quantifies the connectivity in a node’s immediate neighbourhood. At the mesoscale and global levels, modularity captures the extent to which a network can be divided into distinct subnetworks or modules, and system segregation captures the extent to which systems within a functional network are distinctly partitioned 35 . A coarse proxy for system segregation is within-system connectivity.
Studies of prenatal development show that a segregated network structure is present even in utero, with modular subnetworks that coarsely resemble those found in adults 50 , 51 . Inter-regional variation in the width of time windows of synaptogenesis during prenatal and early postnatal development (for example, as seen in ref. 11 ) gives rise to the highly connected hub nodes and modular structure seen in adult brain networks 52 , 53 . Similarly to structural brain development 11 , 12 , functional subnetworks underlying sensory systems become established at an earlier age than do the subnetworks underlying association systems 54 , 55 . Mesoscale segregation increases with age later in childhood and adolescence, probably reflecting the refinement of network architecture; higher-order association systems in particular become more segregated with development 49 , 56 (although some studies do not find positive associations between age and segregation during adolescence, perhaps owing to differences in age range and node or edge definitions; see ref. 57 ). Maturation at the cellular level probably gives rise to these macroscale developmental changes. Inhibitory interneurons have a role in limiting resting-state functional connectivity and establishing the boundaries between brain regions that are necessary for network segregation 58 . In addition, connection strength is associated with microscale properties of connected brain regions, including the size and complexity of layer III pyramidal neurons 59 , 60 , cytoarchitectonic similarity 61 and excitatory–inhibitory receptor balance 62 .
Only a few studies have examined associations between SES and functional brain development using a network science approach (Fig. 2 ), and these studies have used different measures of segregation. Although the use of different measures of segregation at different scales makes an overarching pattern difficult to interpret, here we draw upon existing studies to sketch a theoretical model for future work to detail. One study 63 of infants less than 1-year-old found marginally significant associations between higher SES and both similarity to adult systems and within-system connectivity, a proxy for system segregation. The study’s authors interpret these observations as indicative that greater maturation is associated with higher SES. However, the significant associations were found only at 6 months of age and not at the other time points examined (1, 3, 9 or 12 months). In another study, youth aged 8–22 years from high-SES neighbourhoods show a stronger association between age and local segregation — clustering — than did youth from low-SES neighbourhoods 64 . Although the study authors also examined a mesoscale measure of segregation, namely modularity, the moderating effect of SES on associations between age and modularity was accounted for by local segregation, suggesting that the fundamental driver was variation in local network topology. Specifically, during late childhood, youth from high-SES neighbourhoods showed lower local cortical functional segregation than did youth from low-SES neighbourhoods. However, youth from high-SES neighbourhoods showed a steeper positive relationship between segregation and age during adolescence, such that by their early 20s, they showed greater functional network segregation than youth from low-SES neighbourhoods. Another study of individuals in a similar age range (6–17 years) revealed an interaction between household and neighbourhood SES, such that among youth in low-SES neighbourhoods, higher household SES is associated with greater local functional network segregation (assessed by the clustering coefficient) in the prefrontal cortex 65 . The available evidence is consistent with the hypothesis that higher SES is associated with more protracted functional network development, with youth from high-SES backgrounds showing more widespread connectivity and thus lower segregation early in development, before the rapid development of a more segregated network architecture that continues into adulthood 10 , 11 .
Fig. 2: Associations between socio-economic status and functional brain network segregation.
Trajectories shown in solid and dashed grey lines are conceptual, based on findings interpolated across multiple studies. Horizontal grey lines represent the age ranges of individual studies, as shown on the horizontal axis (grey line 1, ref. 63 ; grey line 2, ref. 64 ; grey line 3; ref. 65 ; grey line 4, ref. 196 ). Brain regions shown in red indicate socio-economic status (SES)-associated differences in functional network segregation, with adolescents from higher-SES backgrounds showing stronger positive associations between age and segregation. Curves are drawn to be consistent with functional network segregation across the studies shown; the studies used a range of measures of segregation, as illustrated in the bottom-right inset. The top-right inset illustrates a common metric of functional connectivity used to estimate functional brain networks: the Pearson product-moment correlation coefficient. Brain images in the lower part of the figure adapted with permission from ref. 64 , OUP.
Full size image
In sum, these studies suggest that the effects of SES on structural development may be reflected in functional development, such that the extended period of structural development associated with high SES gives rise to a longer, slower trajectory of functional network segregation during development, leading to greater segregation. Although longitudinal studies with consistent measures of functional network organization necessary to strictly test these hypotheses do not yet exist, we draw upon existing work to sketch a theoretical model for future work. Lower SES is associated with faster thinning and blunted functional remodelling during childhood and adolescence. In late adolescence and young adulthood, individuals from higher-SES backgrounds show greater cortical thickness and greater segregation than do individuals from lower-SES backgrounds, perhaps as a result of differences in maturation rate. The findings described above also suggest that associations between SES and functional network segregation might follow a progression from local to global across the lifespan, with associations in childhood and early adolescence evident at the local level, and associations at the mesoscale and global level visible later in life. However, more work is needed to understand whether there are truly differing associations at different scales, as few studies thus far have examined multiple measures of segregation in conjunction.
We now turn to two of the most well-studied putative mechanisms underlying SES-associated differences in brain development: stress and cognitive enrichment 5 , 66 . Previous conceptual models have organized variation in early experiences along dimensions of threat (similar to stress) and deprivation (the opposite of cognitive enrichment) 67 , 68 , 69 . We review these factors as possible contributors to the effects of SES on the pace of brain development.
Box 1 Environmental effects on white matter development
If lower socio-economic status (SES) is associated with accelerated brain maturation, we would expect to see differences in the pace of brain maturation reflected in diffusion-based measures of white matter; however, few studies have examined this topic. Typically, studies examining white matter tend to consider fractional anisotropy (FA): the degree of restricted diffusion in a principal direction (λ1) compared with orthogonal directions (λ2 and λ3; see the figure). FA is generally interpreted as a measure of the integrity of a white matter fibre tract. Streamline count is a measure of how many ‘fibres’ can be reconstructed between two brain regions 213 . Structural brain networks can be constructed from measures of regional streamline count or diffusion scalar values averaged along a tract, such as FA.
FA increases steeply in the first few years of life and then more slowly throughout childhood and adolescence 214 , 215 . In developmental studies, group differences in white matter integrity between children from high-SES backgrounds and children from low-SES backgrounds have been identified across various ages, consistently showing that higher SES is associated with higher FA in early childhood (4–7 years) 216 , in middle childhood (8–10 years) 217 , through adolescence (6–19 years and 17–23 years) 107 , 218 and into young adulthood (18–27 years) 219 . Children from higher-SES environments show higher global efficiency of their structural brain networks, indicating that their white matter has many short paths between regions, suggestive of relatively greater integration than in networks of children from lower-SES backgrounds 220 . However, none of these developmental studies examined age–SES interactions. Importantly, measures of FA are related to both axon coherence (compact bundling of several axons in a similar orientation) and myelination, and may also conflate experience-expectant (age-related) myelination with experience-dependent myelination, impairing our ability to detect environmental influences on the rate of maturation.
Lower SES is consistently associated with greater chronic stress 70 , and prior work extensively reviewed the links between SES and multiple conceptualizations of stress 68 , 71 , 72 , 73 , 74 , 75 . There are at least three mechanisms by which chronic stress exposure could accelerate brain development. The first is that repeated use of stress-detection and stress-regulation circuitry, including the amygdala and medial prefrontal cortex, could lead to faster maturation of that circuitry 76 , 77 . The second is that stress could cause faster ageing of the entire body by increasing glucocorticoid levels and allostatic load (physiological wear and tear) and by promoting activation of inflammatory processes 78 . These same physiological processes can be activated by other experiences associated with lower SES, including exposure to environmental toxins (such as lead or air pollution) 79 , poorer sleep quality 80 and less opportunity for physical activity 81 , 82 , 83 , 84 . Stress is associated with accelerated cellular ageing, marked by changes in epigenetic processes such as methylation 85 , 86 , which are detectable in childhood 87 , 88 . Individuals from lower-SES backgrounds tend to enter puberty earlier, and this effect is driven most strongly by experiences of threat 89 , 90 , 91 , 92 . Earlier puberty in turn might also accelerate brain maturation. One study found that the expression of the genes encoding the glucocorticoid receptor and the androgen receptor explained the most variance in cortical thinning in low-income female adolescents living in high-inequality neighbourhoods, suggestive of links between stress and both accelerated puberty and cortical thinning 24 . A third possible mechanism by which chronic stress may accelerate brain development is that young individuals process threat as an overall signal of lack of protection and support — that is, they receive cues that the environment requires maturity — and this triggers adaptive top-down processes that cause development to proceed more quickly. This was recently termed the ‘developmental support hypothesis’ (see ref. 93 ), and aligns with much evolutionary life-history research, including cross-species findings that parental investment is associated with slower maturation 93 , 94 , 95 . Understanding which, if any, of these mechanisms affect the pace of brain development is essential for determining when and how it might be possible to intervene.
Animal models of early-life stress allow us to address issues of causality that cannot be examined in humans. The animal paradigm most analogous to the economic deprivation and stress associated with SES is the limited bedding and nesting model in rodents, which involves limiting the dam’s access to sufficient bedding and nesting material. Although the limited bedding paradigm fails to capture many of the social, emotional and cognitive aspects of being raised in a low-SES environment, this constraint does result in fragmented and unpredictable nurturing behaviours and increased glucocorticoid release in the pups 96 , 97 . Offspring of the dams exposed to this paradigm show earlier declines in the levels of markers of postnatal neurogenesis in the hippocampus, earlier increases in the levels of markers of synaptic maturity, earlier increases in the level of myelin basic protein and impairments in cognitive function 98 , 99 , 100 . They also show an initial increase in neuronal proliferation in the hippocampus in early life, but at later times show reduced numbers of neurons and reduced hippocampal volume, suggestive of an earlier peak in neurogenesis 101 . Prefrontal areas and the hippocampus show reduced spine density following exposure to this paradigm in the early postnatal period. These changes are associated with impairments in cognitive function 102 , 103 that are prevented by blocking the effect of corticotropin-releasing hormone, a stress-linked neuropeptide, immediately following exposure to this paradigm. This finding is consistent with a large body of work showing that some effects of the early environment are modulated by glucocorticoids 85 , 104 . We now turn to the question of whether SES differences in cognitive enrichment or deprivation also drive differences in the pace of brain development.
Exposure to a complex environment with a variety of experiences and diverse learning materials is known as cognitive enrichment. The absence of cognitive enrichment is considered deprivation 67 , 68 . Children growing up in higher-SES homes tend to be exposed to more complex and cognitively stimulating environments 105 , and cognitive enrichment is associated with improved cognition in youth independent of stress exposures 69 , 106 , 107 , 108 . In one study, cognitive stimulation also mediated associations between SES and cortical thickness in prefrontal areas 107 , highlighting its potential role as a mechanism of the influence of SES on brain development in childhood. Recapitulating these findings, SES-associated differences in children’s cognitive function have been reported to be mediated by cognitive enrichment in the home 109 . Some models suggest that the absence of cognitive enrichment in specific domains leads to accelerated synaptic pruning in brain regions that process complex cognitive and social stimuli 67 , 68 . The converse of this argument is that specific cognitive inputs might delay synaptic pruning in relevant brain circuitry.
As in studies of stress, animal models allow us to investigate the causal influence of cognitive enrichment on brain development. Environmental enrichment paradigms typically have two main components: novel objects and novel social partners. Environmental enrichment in both juvenile and adult animals has been shown to lead to increased cortical thickness 110 , 111 , driven by increases in dendritic volume and branching 112 , 113 , dendritic spine count 112 , 114 , synaptogenesis and glial proliferation 115 , 116 (reviewed in ref. 117 ). As little as 4 days of enrichment produces measurable changes in cortical thickness in rodents 118 , 119 , and longer exposure is associated with longer retention of increased thickness after return to a standard environment 120 . Enrichment may also affect cortical surface area, but it is not commonly measured 121 . Increased synaptogenesis, glial proliferation and dendritic plasticity could indicate a prolonged period of maturation leading to more complex brain circuitry, as computational models that suggest early synaptic overgrowth and overall slower development are advantageous for adult network abilities 32 , 122 . In sum, there is some evidence that children’s early experiences of stress and cognitive enrichment influence the pace of brain development.
Consequences for plasticity
Understanding how children’s experiences affect the pace of brain maturation has consequences for understanding brain plasticity. Brain plasticity can be conceptualized in two ways: as a process and as a potential. The process of brain plasticity, including long-term potentiation and other structural and functional changes in response to experience, occurs throughout life. However, the brain’s plasticity as potential for change varies with age. Developmental processes, including myelination, inhibition and the formation of perineuronal nets (PNNs; lattice-like extracellular structures that enwrap neurons and act as a physical brake on plasticity) decrease the brain’s ability to change as children get older 123 , 124 (Box 2 ). If brain development proceeds more quickly in children from low-SES backgrounds, windows of high plasticity could also close more quickly in these children, reducing the brain’s sensitivity to future experiences. In this section, we review evidence from animal models that experiences of stress and cognitive enrichment affect plasticity. Most of this research was done in adult animals, but the results suggest that these experiences would affect plasticity during development as well.
Studies in animal models have broadly shown that early-life stress decreases synaptic plasticity and promotes the developmental processes that reduce plasticity (such as inhibition and myelination). Offspring of dams exposed to the limited-bedding paradigm show earlier increases in the levels of markers of synaptic maturity, earlier increases in the level of myelin basic protein, an increased number of PNNs and reductions in adult synaptic plasticity, accompanied by impairments in cognitive function, compared with control offspring 98 , 99 , 101 , 125 . The limited-bedding paradigm also causes reduced spine plasticity in the offspring’s prefrontal cortex and hippocampus 102 , 103 . Increased myelination is not always observed following early-life stress: one study found that early social isolation leads to a decrease in myelination in the prefrontal cortex 126 . Therefore, the impact of stress on myelination may depend on the type of stressor and the brain area examined. There are also indirect links between early-life stress and plasticity: early-life stress accelerates pubertal timing (age of onset of pubertal development or age at menarche), and ovarian hormones increase cortical inhibition 92 , 127 , 128 . Studies of stress on plasticity in humans are rare. One study of post-mortem brains found that individuals who were exposed to child abuse had increased numbers of mature myelinating oligodendrocytes in the ventral medial prefrontal cortex 129 . Another study used neuroimaging to show that veterans with post-traumatic stress disorder had higher T1-weighted/T2-weighted MRI signal, a marker of myelination, in the hippocampus 130 . Both studies are consistent with the hypothesis that stress increases myelination, and may thereby limit plasticity.
Environmental-enrichment paradigms prolong and enhance plasticity. Enrichment during the juvenile period decreases the number of PNNs 131 , enhances synaptic plasticity in the form of long-term potentiation and depression 132 and influences parvalbumin-positive neuron expression 131 , 133 , 134 . In adulthood, enrichment keeps inhibition at juvenile levels, prolonging early periods of plasticity 135 , 136 , 137 . Enrichment paradigms can also enhance plasticity in adults long past juvenile critical periods 138 by reducing inhibition 137 , 139 , decreasing PNN stability 139 , 140 , 141 or increasing myelin remodelling 142 , all potent contributors to plasticity. Environmental enrichment increases neuronal secretion of the cytokine interleukin-33 (IL-33), which signals to microglia to engulf PNNs, increasing synaptic plasticity 141 . Environmental enrichment also enhances levels of neurotransmitters, including noradrenaline, dopamine and serotonin, which increase cortical plasticity and facilitate cortical remodelling 143 , 144 , 145 , 146 . Mice lacking the dopamine D2 receptor or the dopamine D4 receptor fail to benefit in longevity from enriched environments 147 , 148 . The social interaction component of an enriched environment increases release of oxytocin, which enhances plasticity 149 and protects against stress-related changes in plasticity 150 , 151 . To our knowledge, studies examining the impact of cognitive enrichment on plasticity in humans do not yet exist.
Stress and cognitive enrichment broadly capture the valence of experiences: stressful experiences are negative and should be avoided, whereas environmental enrichment paradigms are designed to be positive and rewarding. However, valence is not the only salient property of such experiences. The timing of experiences also has implications for plasticity (Fig. 3 ). Repeated exposure to the same experience should signal that the experience is more likely to occur consistently in the future, and that the brain should optimize to respond to it, even at a cost to plasticity. Experience-dependent myelination and PNN formation are two potential mechanisms by which repeated activation of brain circuitry might lead to reduced plasticity 124 .
Fig. 3: Integrative theory: childhood experiences affect the pace of brain development.
According to our model, experiences that are chronic or repetitive and negative encourage faster maturation and increase allostatic load, potentially restricting plasticity. Experiences that are rare and positive, triggering surprise and awe, are associated with strong neurochemical signals to delay maturational processes and enhance plasticity. Experiences in the other quadrants (rare and negative, or repetitive and positive) are predicted to have smaller effects on the global pace of maturation.
Magnetic resonance spectroscopy, glutamate chemical exchange saturation transfer, GABA chemical exchange saturation transfer
Extracellular matrix organization (including perineuronal nets)
Multicompartment diffusion imaging (for example, neurite orientation dispersion and density imaging or soma and neurite density imaging)
Fractional anisotropy or mean diffusivity from diffusion imaging; multicompartment diffusion imaging; T1-weighted to T2-weighted ratio; magnetization transfer; quantitative MRI
Levels of neurotransmitters (such as dopamine, acetylcholine or serotonin)
Positron emission tomography, functional MRI or resting-state functional MRI of neuromodulatory nuclei
Future directions and conclusions
In this Perspective, we have considered evidence that experiences associated with childhood SES affect not only the outcome but also the pace of brain development, with potential influences on brain plasticity throughout life. We argue that low exposure to stress and high exposure to novel positive experiences promote protracted structural brain development, which gives rise to a later, longer trajectory of functional network segregation, ultimately leading to more efficient cortical networks in adulthood.
However, this model is based on incomplete data. Studies to date have not been fully representative of human diversity, focusing primarily on Western populations with nutritional excess 168 , 169 , 170 . Studies have also been limited by methodological challenges, cross-sectional samples, lack of connection to adult research and correlational designs. Below, we discuss promising approaches to overcome these limitations and directly test our hypotheses in future research.
Methodological advances are necessary to fully understand how early experiences affect the pace of brain development. The application of network methods to developmental data is still in its infancy, as researchers take on the challenge of describing nodes and edges of brain networks in a biologically accurate and meaningful way 171 , 172 . Studies have used many different measures of segregation to characterize functional brain networks, and it will be crucial for future research to examine how different measures relate to each other and to SES over development. The field has also become increasingly aware of how methodological decisions, including correcting for head motion 173 , 174 , 175 , 176 and physiological artefacts 177 , 178 , 179 , affect study conclusions, and thus affect our ability to make inferences across sets of studies. New methods are also needed for integrating structural and functional brain data. Few articles have examined both functional and structural brain development in the context of SES, and little is known about the relative ordering of trajectories of cortical thinning, white matter development and functional network segregation. Recent work has attempted to link changes in structure to changes in function 54 , 180 , but the sequence of developmental progression, let alone environmental influences on that sequence, remains murky. Another area for future work involves linking histology and electrophysiology data to structural and functional MRI findings in animal models to facilitate translation to human work. Such an effort would enable us to test how early-life experiences influence cellular developmental processes, including myelination and inhibition, that give rise to macro-level measures, including cortical thickness, surface area and network segregation.
Many of the studies reviewed herein are cross-sectional. Cross-sectional data have inherent limitations when developmental processes are being examined, foremost among them the inability to infer the shape of developmental trajectories 9 . Cross-sectional studies cannot establish temporal precedence and, if sampling is non-random, associations with age may be driven by the characteristics of the sample rather than by age 181 , 182 . Longitudinal studies, such as the Adolescent Brain Cognitive Development (ABCD) study 183 , 184 and the upcoming HEALthy Brain and Child Development (HBCD) study 185 , will be necessary to fully understand how early environments influence trajectories of functional and structural brain development 182 , 186 . Data from these longitudinal studies will enable us to examine whether changes in brain structure correspond to changes in functional network segregation, and whether measures of the early environment predict earlier or later peaks in these trajectories.
An important future direction is determining whether SES effects on early brain maturation set the stage for early brain ageing 3 , 187 , 188 . There is initial evidence from a prospective study that traces of childhood SES are still present in the brain structure of young adults aged 23–25 years, even when adulthood SES is controlled for 189 . We do not yet know whether this is also true of older adults, but studies suggest that cognitive enrichment might be important: cognitive stimulation in childhood is associated with larger brain volumes 190 and better cognition in old age 191 when adulthood SES is controlled for. Furthermore, a longitudinal study showed that higher levels of early cognitive stimulation are associated with slower cognitive decline and less neuropathology with ageing 192 . Studies examining adulthood SES and brain structure and function find results that are broadly consistent with the theoretical framework we outline in this Perspective 193 , 194 , 195 , 196 . In one study, adults from higher-SES backgrounds showed a weaker negative association between segregation and age than did adults from lower-SES backgrounds, consistent with an interpretation of a slower decline in functional network organization in higher-SES adults 197 . Associations with adulthood SES were stronger than associations with childhood SES; however, because adulthood SES and childhood SES are correlated, these factors can be difficult to disentangle. Childhood SES is difficult to measure in an ageing sample because of recall biases 198 . Relationships between parental education and childhood experiences may also have been different when today’s 80-year-olds were children from how they are for today’s 6-year-olds, making retrospective report of SES in adults difficult to map to current developmental research. In addition, low-SES populations may be poorly represented in ageing research, owing to lower-life expectancy 187 and higher prevalence of other diseases and health issues that would exclude these populations from studies of healthy ageing 3 , 188 . Ideally, future studies will follow individuals from birth to old age, although this may be more feasible in animal models (for example, as in ref. 199 ).
Although longitudinal observational studies are useful, intervention studies are necessary to directly test whether children’s early experiences cause slower or faster brain development. Future work should test whether cognitive enrichment in humans leads to changes in the pace of brain development, and whether the timing of enrichment influences these effects. Although we cannot evaluate the impact of creating early stressful experiences for children, we can learn from the effect of naturally occurring stressors. The emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) as a global public health crisis has resulted in an unforeseen natural experiment on how the timing — that is, the age of children when it occurred — and the severity of a stressor affect the pace of children’s maturation. However, the effect of the stress has been non-random, as the crisis has disproportionately affected lower-income communities and people from minority racial or ethnic groups and other marginalized populations 200 , 201 . Following the brain development of children who lived through this period will yield insight into the importance of stress timing on the rate of maturation.
It is possible to investigate causal effects of cognitive enrichment by studying educational interventions. Education is broadly beneficial for children’s development, leading to increased cognitive ability, and better health and wellness throughout life 202 , 203 . Both the type and the timing of education could influence brain plasticity. Rote practice is likely to drive faster maturation of the brain systems involved, which would be beneficial for the task practised (for example, reading and writing), but it could compromise the ability to learn novel tasks. By contrast, rich and varied experiences that capture children’s attention and enhance their motivation, boosting levels of acetylcholine and dopamine, could decelerate the rate of brain maturation. Our model also predicts that educational experiences earlier in childhood will have a bigger effect on brain development and plasticity than experiences later in childhood, by changing the trajectory of maturation. Evidence for the efficacy of early interventions, such as from the Abecedarian Project and the Perry Preschool Program, is broadly consistent with this hypothesis 204 , 205 ; however, direct comparisons of the same curricula at different ages are rare, and thus the neural outcomes of changing the timing of such interventions are not yet known. Determining the consequences of educational strategies for the pace of brain maturation is an important area of future research.
In conclusion, disparate strands of evidence from neuroscience, psychology and medicine are consistent with a model in which the early environment affects not only the outcome but also the pace of human brain development. We propose that high stress and low cognitive enrichment accelerate developmental changes in cortical thickness and surface area, and shift the trajectory and amplitude of functional network segregation across development. We argue that changes in the pace of brain development also affect plasticity during development. Our work provides a generative theoretical framework for research on links between childhood experiences and brain changes over the lifespan, and reinforces the pressing need to elucidate changes in early development that lead to disparities in later-life outcomes. If we can develop new screening tools to detect accelerated development, we will be better able to implement educational and other interventions earlier, and prevent cascading consequences of early maturation for mental and physical health.
Citation diversity statement
Recent work in neuroscience and other fields has identified a bias in citation practices such that articles authored by women and scholars from minority racial or ethnic groups are undercited relative to the number of such articles in the field 206 , 207 , 208 . The expected gender proportions in reference lists of five top neuroscience journals as reported by Dworkin et al. were 6.7% for woman (first author)/woman (last author), 9.4% for man/woman, 25.5% for woman/man and 58.4% for man/man 209 . Inclusion of citation diversity statements has been proposed as a way of increasing transparency surrounding citation practice 210 , 211 . We obtained the predicted gender of the first and last authors of each reference by using databases that store the probability of a name belonging to a woman or man, and classifying as ‘unknown’ any names with under 70% predicted accuracy 209 , 212 . Excluding self-citations of the authors of the current Perspective, for our references the proportions are 26% woman (first author)/woman (last author), 20% man/woman, 21% woman/man and 33% man/man. This method is limited in that the databases used may not, in every case, be indicative of gender identity, and in that it does not account for intersex, non-binary or transgender people. In addition, the expected proportions above were calculated across all neuroscience subfields, and may differ for particular subfields, such as developmental neuroscience. We look forward to future work to better understand how to support equitable practices in science.
Duncan, G. J., Magnuson, K., Kalil, A. & Ziol-Guest, K. The importance of early childhood poverty. Soc. Indic. Res. 108, 87–98 (2012).